Background: Detection of atrial fibrillation (AF) after ischemic stroke is important because anticoagulation is indicated to reduce the risk of recurrent stroke. However, no consensus exists about the optimum method for detecting underlying paroxysmal AF not apparent on presentation with stroke. The aim of this study was to characterize the rate, timing, and predictors of delayed detection of AF after stroke. Methods: The Virtual International Stroke Trials Archive provided data from 3464 patients in the placebo arms of 4 clinical trials of therapies for acute ischemic stroke. Patients who had AF by history or on the baseline electrocardiogram were excluded. Electrocardiograms were obtained routinely and as clinically indicated. The time to detection of AF was evaluated using Kaplan-Meier survival statistics. Cox proportional hazards analysis was used to evaluate risk factors for AF. Results: Among 2504 qualifying patients, AF was detected in 174 (6.9%; 95% confidence interval [CI] 6.0%-8.0%). In 68% of patients, AF was detected more than 48 hours after presentation. Detection of AF was associated with increasing age (hazard ratio [HR] 1.6/decade; 95% CI 1.4-1.9; P < .005), female sex (HR 1.7; CI 1.2-2.4; P < .005), congestive heart failure (HR 1.9; CI 1.1-3.4; P = .02), and the absence of hypertension (HR 1.6; CI 1.1-2.2; P = .01). Conclusions: Delayed detection of AF was common in this large cohort of patients carefully monitored after ischemic stroke. Current methods of screening may fail to detect underlying paroxysmal AF in a substantial proportion of patients.